Development and external validation of two nomograms to predict overall survival and occurrence of distant metastases in adults after surgical resection of localised soft-tissue sarcomas of the extremities: a retrospective analysis

Background: The current American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) staging system does not have sufficient details to encompass the variety of soft-tissue sarcomas, and available prognostic methods need refinement. We aimed to develop and externally validate two prediction nomograms for overall survival and distant metastases in patients with soft-tissue sarcoma in their extremities.

Methods: Consecutive patients who had had an operation at the Istituto Nazionale Tumori (Milan, Italy), from Jan 1, 1994, to Dec 31, 2013, formed the development cohort. Three cohorts of patient data from the Institut Gustave Roussy (Villejuif, France; from Jan 1, 1996, to May 15, 2012), Mount Sinai Hospital (Toronto, ON, Canada; from Jan 1, 1994, to Dec 31, 2013), and the Royal Marsden Hospital (London, UK; from Jan 1, 2006, to Dec 31, 2013) formed the external validation cohorts. We developed the nomogram for overall survival using a Cox multivariable model, and a Fine and Gray multivariable model for the distant metastases nomogram. We applied a backward procedure for variables selection for both nomograms. We assessed nomogram model performance by examining overall accuracy (Brier score), calibration (calibration plots and Hosmer-Lemeshow calibration test), and discrimination (Harrell C index). We plotted decision curves to evaluate the clinical usefulness of the two nomograms.

Findings: 1452 patients were included in the development cohort, with 420 patients included in the French validation cohort, 1436 patients in the Canadian validation cohort, and 444 patients in the UK validation cohort. In the development cohort, 10-year overall survival was 72·9% (95% CI 70·2-75·7) and 10-year crude cumulative incidence of distant metastases was 25·0% (95% CI 22·7-27·5). For the overall survival nomogram, the variables selected applying a backward procedure in the multivariable Cox model (patient's age, tumour size, Fédération Française des Centres de Lutte Contre le Cancer [FNCLCC] grade, and histological subtype) had a significant effect on overall survival. The same variables, except for patient age, were selected for the distant metastases nomogram. In the development cohort, the Harrell C index for overall survival was 0·767 (95% CI 0·743-0·789) and for distant metastases was 0·759 (0·736-0·781). In the validation cohorts, the Harrell C index for overall survival and distant metastases were 0·698 (0·638-0·754) and 0·652 (0·605-0·699; French), 0·775 (0·754-0·796) and 0·744 (0·720-0·768; Canadian), and 0·762 (0·720-0·806) and 0·749 (0·707-0·791; UK). The two nomograms both performed well in terms of discrimination (ability to distinguish between patients who have had an event from those who have not) and calibration (accuracy of nomogram prediction) when applied to the validation cohorts.

Interpretation: Our nomograms are reliable prognostic methods that can be used to predict overall survival and distant metastases in patients after surgical resection of soft-tissue sarcoma of the extremities. These nomograms can be offered to clinicians to improve their abilities to assess patient prognosis, strengthen the prognosis-based decision making, enhance patient stratification, and inform patients in the clinic.

Funding: None.